Configural processing as an optimized strategy for robust object recognition in neural networks
Abstract Configural processing, the perception of spatial relationships among an object’s components, is crucial for object recognition, yet its teleology and underlying mechanisms remain unclear. We hypothesize that configural processing drives robust recognition under varying conditions. Using ide...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-03-01
|
| Series: | Communications Biology |
| Online Access: | https://doi.org/10.1038/s42003-025-07672-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849762238315888640 |
|---|---|
| author | Hojin Jang Pawan Sinha Xavier Boix |
| author_facet | Hojin Jang Pawan Sinha Xavier Boix |
| author_sort | Hojin Jang |
| collection | DOAJ |
| description | Abstract Configural processing, the perception of spatial relationships among an object’s components, is crucial for object recognition, yet its teleology and underlying mechanisms remain unclear. We hypothesize that configural processing drives robust recognition under varying conditions. Using identification tasks with composite letter stimuli, we compare neural network models trained with either configural or local cues. We find that configural cues support robust generalization across geometric transformations (e.g., rotation, scaling) and novel feature sets. When both cues are available, configural cues dominate local features. Layerwise analysis reveals that sensitivity to configural cues emerges later in processing, likely enhancing robustness to pixel-level transformations. Notably, this occurs in a purely feedforward manner without recurrent computations. These findings with letter stimuli successfully extend to naturalistic face images. Our results demonstrate that configural processing emerges in a naíve network based on task contingencies, and is beneficial for robust object processing under varying viewing conditions. |
| format | Article |
| id | doaj-art-61951b39e5da45cc8fdd76ec61fbc232 |
| institution | DOAJ |
| issn | 2399-3642 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Communications Biology |
| spelling | doaj-art-61951b39e5da45cc8fdd76ec61fbc2322025-08-20T03:05:48ZengNature PortfolioCommunications Biology2399-36422025-03-018111110.1038/s42003-025-07672-1Configural processing as an optimized strategy for robust object recognition in neural networksHojin Jang0Pawan Sinha1Xavier Boix2Department of Brain and Cognitive Engineering, Korea UniversityDepartment of Brain and Cognitive Sciences, Massachusetts Institute of TechnologyArtificial Intelligence Laboratory, Fujitsu Research of AmericaAbstract Configural processing, the perception of spatial relationships among an object’s components, is crucial for object recognition, yet its teleology and underlying mechanisms remain unclear. We hypothesize that configural processing drives robust recognition under varying conditions. Using identification tasks with composite letter stimuli, we compare neural network models trained with either configural or local cues. We find that configural cues support robust generalization across geometric transformations (e.g., rotation, scaling) and novel feature sets. When both cues are available, configural cues dominate local features. Layerwise analysis reveals that sensitivity to configural cues emerges later in processing, likely enhancing robustness to pixel-level transformations. Notably, this occurs in a purely feedforward manner without recurrent computations. These findings with letter stimuli successfully extend to naturalistic face images. Our results demonstrate that configural processing emerges in a naíve network based on task contingencies, and is beneficial for robust object processing under varying viewing conditions.https://doi.org/10.1038/s42003-025-07672-1 |
| spellingShingle | Hojin Jang Pawan Sinha Xavier Boix Configural processing as an optimized strategy for robust object recognition in neural networks Communications Biology |
| title | Configural processing as an optimized strategy for robust object recognition in neural networks |
| title_full | Configural processing as an optimized strategy for robust object recognition in neural networks |
| title_fullStr | Configural processing as an optimized strategy for robust object recognition in neural networks |
| title_full_unstemmed | Configural processing as an optimized strategy for robust object recognition in neural networks |
| title_short | Configural processing as an optimized strategy for robust object recognition in neural networks |
| title_sort | configural processing as an optimized strategy for robust object recognition in neural networks |
| url | https://doi.org/10.1038/s42003-025-07672-1 |
| work_keys_str_mv | AT hojinjang configuralprocessingasanoptimizedstrategyforrobustobjectrecognitioninneuralnetworks AT pawansinha configuralprocessingasanoptimizedstrategyforrobustobjectrecognitioninneuralnetworks AT xavierboix configuralprocessingasanoptimizedstrategyforrobustobjectrecognitioninneuralnetworks |